ISP speed tests

The ubiquity of speedtest.net

Almost all speed tests provided by internet service providers are based on Ookla's speedtest.net.

For example, from the internet service providers mentioned at ISP Hosted Internet Speed Tests, AT&T, Cablevision, CCI, CenturyLink, Charter, Consolidated Communications, Cox, Fios, Frontier, GCI, Grande Communications, Midcontinent, Optimum, Quest, Rise Broadband, Shaw, Skybeam, Spectrum, SureWest, TDS, Telus, Time Warner, US Internet, Verizon and Wave all use Ookla's Speed test.

The exceptions are Comcast and Xfinity (both use Xfinity xFi Speed Test), Google Fiber and RCN and maybe CableOne too (which requires a login).

So at least 26 of the 31 of these internet service providers use Ookla's Speedtest. That is 84%. That isn't that strange if you know that everyone is able to create a customized speed test in a few steps.

Testing all modified versions of Ookla's speednet.net probably makes no sense. To be on the safe side, it is measured whether the download speeds differ significantly.

ISP Speedtest versus speedtest.net

The download speed is measured alternately with the ISP speedtest and speedtest.net. Speedtest.net uses a test server that matches the test server of the ISP speedtest as much as possible. First the test is done with Ziggo. Ziggo offers a download speed of 50Mbps.

Ziggo

Tested with Firefox the following measurements are done.

Test Server: Ziggo, Utrecht (fixed it is not possible to choose another test server)

  1. 48.9Mbps (19 seconds)
  2. 49.9Mbps (18 seconds)
  3. 48.7Mbps (18 seconds)
  4. 49.5Mbps (18 seconds)
  5. 50.0Mbps (18 seconds)

That gives an average of 49.4Mbps and a standard deviation of 0.6Mbps for the downloadspeed. The average and standard deviation for the time required are respectively 18 seconds and 0 seconds.

Speedtest.net

Tested with Firefox the following measurements are done.

Test Server: Korton Group BV - Hoofddorp (suggested test server, it is not possible to select a test server in Utrecht)

  1. 49.51Mbps (25 seconds)
  2. 47.32Mbps (26 seconds)
  3. 49.92Mbps (26 seconds)
  4. 49.29Mbps (26 seconds)
  5. 49.89Mbps (26 seconds)

That gives an average of 49.2Mbps and a standard deviation of 1.08Mbps for the downloadspeed. The average and standard deviation for the time required are respectively 26 seconds and 0 seconds.

There is no significant difference between these two tests for the download speed. The Ziggo speed test is significantly faster than speedtest.net.

Midcontinent

Tested with Firefox the following measurements are done.

Test Server: Grand Forks, ND (fixed it not possible to choose another test server)

  1. 49.3Mbps (23 seconds)
  2. 46.5Mbps (22 seconds)
  3. 48.5Mbps (22 seconds)
  4. 49.1Mbps (22 seconds)
  5. 48.3Mbps (22 seconds)

That gives an average of 48.3Mbps and a standard deviation of 1.1Mbps for the downloadspeed. The average and standard deviation for the time required are respectively 22 seconds and 0 seconds.

Speedtest.net

Tested with Firefox the following measurements are done.

Test Server: Grand Forks, ND (this test server is manually chosen)

  1. 49.37Mbps (29 seconds)
  2. 48.58Mbps (29 seconds)
  3. 48.72Mbps (29 seconds)
  4. 49.53Mbps (29 seconds)
  5. 48.98Mbps (29 seconds)

That gives an average of 49.0Mbps and a standard deviation of 0.41Mbps for the downloadspeed. The average and standard deviation for the time required are respectively 29 seconds and 0 second.

There is no significant difference between these two tests for the download speed. The Midcontinent speed test is significantly faster than speedtest.net.

No significant differences

According to expectations, there are no significant differences in the measured download speed between speedtest.net and an ISP branded version of the Ookla Speedtest.

Significant differences

The ISP branded versions of speedtest.net are significantly faster than the original speedtest. Note that this difference cannot be explained by the location of the test server.

If your ISP offers a speed test based on the Ookla speed test, it is best to use it.